LucasArts engineer Marco Salvi has posted an in-depth explanation of his method for HDR rendering, prompted by a SIGGRAPH 2006 lecture series by Valve employees. Salvi first touches on Valve's own HDR process (added to the Source engine in 2005), which he points out runs "with MSAA on relatively old hardware [and] executes tone mapping and MSAA resolve in the proper correct order with no extra performance cost, something that a lot of modern games can't still get right today."

"Through image segmentation techniques," Valve's code "’simply’ tries to determine if the previous frame has been under or over exposed and a new exposure value is adjusted to compensate for problems with the previous frame(s)." Salvi notes that this method doesn't allow for reliably determining average logarithmic luminance, which in his experience has led to "an overall flat and over or under saturated feeling." So he introduces his own plan: "get rid of the exposure search through previous frames feedback and compute it the proper way!"

Salvi's method computes per-pixel logarithmic luminance and outputs that to the alpha channel; he includes a code example of the relevant math, and delves into further specific details.

As a bonus, Valve's Gary McTaggart and Chris Green, both presenters of the lectures that spurred Salvi's post, drop by in the blog's comment section to explain their reasoning behind the route they took for the Source engine's HDR implementation.

Posted by Chris Remo at 10:59 AM | Permalink | Comments (18)

In the latest Intel-sponsored feature, Lightspeed Publishing's Lee Purcell lays out deferred mode image processing, a new addition to the Intel Integrated Performance Primitives Library, which speeds up complex image-processing tasks with up to 3X performance increases.

Purcell begins by laying out current issues in dealing with image processing, which becomes more challenging with the ever-increasing resolution and color demands that come alongside the evolutions in high-definition imaging. Intel's DMIP attempts to address the issue. As software engineer Alexander Kibkalo, quoted by Purcell, explains:

"Fundamentally, DMIP optimizes overall image-processing tasks within an application while individual Intel IP functions can be optimized without requiring knowledge of the environment or the conditions of the function call. DMIP provides the detailed descriptions of each task in DAG form and the appropriate preferences can then be applied for optimizing the routines.

"In terms of parallelization on Intel processors, DMIP tries to maintain a balance of the slice size. Keeping it comparatively small lets the slice fit into L2 cache and enable efficient pipelining. Splitting the slices into comparatively large sub-slices allows them to be efficiently handled by the available processor cores (for example, by 4 lines for a quad-core processor or by 8 lines for an 8-core processor). For achieving the best performance, the actual method of splitting should be tailored to the individual processor on which the application is being run."

Purcell also notes several secondary benefits:

"For example, matching the Intel IP function algorithms to the low-level optimizations for the Intel Streaming SIMD Extensions (Intel SSE) (from Intel SSE through Intel SSE4) can improve overall application performance substantially. Software developers and their application designs benefit from well-established, highly refined algorithms that address many of the most important programming operations. Intel's support network also adds value and utility to these libraries."

You can now read the full feature, which includes detailed information and visual aids.

Posted by Chris Remo at 05:12 PM | Permalink | Comments (16)